FMRGAN: Feature Mapping Reconstruction GAN for Rolling Bearings Fault Diagnosis Under Limited Data Condition

IEEE Sensors Journal, Journal Year: 2024, Volume and Issue: 24(15), P. 25116 - 25131

Published: June 24, 2024

Language: Английский

---

Hierarchical Grey Wolf Optimizer-Tuned Flexible Residual Neural Network with Parallel Attention Module for Bearing Fault Diagnosis

IEEE Sensors Journal, Journal Year: 2024, Volume and Issue: 24(12), P. 19626 - 19635

Published: May 13, 2024

Rolling bearings are vital components of rotating machinery, and their regular operation directly affects the machine lifespan operating status. Aimed at improving accuracy fault diagnosis for rolling bearings, a hierarchical grey wolf optimizer (HGWO)-tuned flexible residual neural network (FResNet) with parallel attention module (PAM) is proposed. Specifically, CNN based designed to form FResNet, which allows changing numbers convolution layers kernels as an iterates. Optimal model structure parameters configured by HGWO non-linear convergence factor position update strategy. On other hand, PAM convolutional fuse output weights channel spatial attention. As result, integration HGWO, PAM, FResNet forms effective bearing diagnosis, named HGWO-PAM-FResNet. Finally, viability efficacy proposed HGWO-PAM-FResNet verified using dataset from Case Western Reserve University, higher compared intelligent models demonstrated under different noise variable load conditions.

Language: Английский

---

Sequential Feature-Augmented Deep Multilabel Learning for Compound Fault Diagnosis of Rotating Machinery With Few Labeled and Imbalanced Data

IEEE Transactions on Industrial Informatics, Journal Year: 2024, Volume and Issue: 20(12), P. 13947 - 13955

Published: Aug. 14, 2024

Accurate fault diagnosis of rotating machinery is essential for smooth and safe operations mechanical systems, various data-driven methods have been developed based on massive sensing data. However, the frequent occurrence compound faults makes it much challenging. Meanwhile, few labeled imbalanced data further complicate design methods. To address these issues, this article proposes a novel sequential feature augmented deep multilabel learning model diagnosis. Specifically, by integrating convolutional neural network with long short-term memory, stacked sparse autoencoder to extract high-dimensional marginal time-sequential features from Then, supervised learn relationships among single finally realize accurate Experimental results demonstrated that our could cope well scenarios outperforms many existing models.

Language: Английский

---

Feature Selection-Based Multiview Concentration for Multivariate Time Series Classification and Its Application

IEEE Sensors Journal, Journal Year: 2023, Volume and Issue: 24(4), P. 4798 - 4806

Published: Dec. 28, 2023

In this article, a feature selection-based multiview concentration (FS-MVC) algorithm is proposed for multivariate time series classification (MTSC). The data-driven excavator fault diagnosis task used as an application case of the MTSC algorithm. three steps FS-MVC comprise multidimensional extraction, concentration, and ensemble. (MTSs) are mapped to multiple spaces by various transformations extract dimension-dependent dimension-independent features. A new pooling extracted, which denotes power proportion positive values (PPPV) in map generated convolution operation. ensemble-group selection framework introduced into remove redundant features generate multiviews through vector concentration. ensemble, each view input corresponding classifier, then, final prediction label obtained hard voting Feature diversity improved via PPPV features, while stability enhanced significantly framework. Finally, superiority over other state-of-the-art algorithms demonstrated both contrast experiments on public UEA MTS practical datasets.

Language: Английский

---

Operation Mode Recognition with Multiple Sensing Data for Electro-Mechanical Actuator based on Deep-shallow Fusion Network

IEEE Transactions on Instrumentation and Measurement, Journal Year: 2024, Volume and Issue: 73, P. 1 - 10

Published: Jan. 1, 2024

In next-generation aircraft, Electro-Mechanical Actuators (EMAs) are increasingly used. But the safety of EMA is not sufficient for primary flight control actuation aircraft. One effective way to improve develop Prognostics and Health Management (PHM). However, variable operation modes make it difficult implement high-performance PHM. Thus, need be recognized, but high similarity sensing data between different making challenging. a new deep-shallow fusion network with convolutional neural network, self-attention mechanism Bayesian (CSBN) proposed mode recognition, which can overcome challenge multiple data. CSBN based recognition method, statistical features firstly extracted discretized. Then, conducted discretized on CSBN. Finally, output used as results. To validate its effectiveness, experiments utilizing practical implemented. Experimental results demonstrate that suitable recognition.

Language: Английский

---